Application value of CyTOF 2 mass cytometer technology at single-cell level in human gastric cancer cells

(1)

Contents lists available atScienceDirect

Experimental Cell Research

journal homepage:www.elsevier.com/locate/yexcr

Application value of CyTOF 2 mass cytometer technology at single-cell level

in human gastric cancer cells

Shuangshuang Han

a,1

_{, Xia Jiang}

b,1

_{, Xiao-Feng Sun}

c

_{, Hong Zhang}

d

_{, Chao Li}

a

_{, Zengren Zhao}

b

_,

Weifang Yu

a,∗

a_{Departments of Endoscopy Center, The First Hospital of Hebei Medical University, 050031, Shijiazhuang, Hebei, China} b_{Departments of General Surgery, The First Hospital of Hebei Medical University, 050031 Shijiazhuang, Hebei, China}

c_{Department of Oncology and Department of Clinical and Experimental Medicine, Linköping University, SE-58183, Linköping, Sweden} d_{School of Medical Sciences, Örebro University, SE-70182, Örebro, Sweden}

A R T I C L E I N F O Keywords: Mass cytometry Gastric cancer Single-cell viSNE Citrus A B S T R A C T

Chemotherapy and radiotherapy are main adjuvant therapies for the treatment of gastric cancer, the treatment effects are individual difference, but the specific mechanism is unknown. CyTOF 2 mass cytometer (CyTOF) enables the detecting up to 135 parameters on single cell, the emergence of which is an opportunity for pro-teomics research. We first tried to apply CyTOF technique to gastric cancer cells. We verified applicability of CyTOF in gastric cancer cells, and analyzed the responses of seventeen proteins to chemoradiotherapy in human gastric cancer AGS cells. To analyze the high dimensional CyTOF data, we used two statistical and visualization tools including viSNE and Citrus. Two specific clusters were found which had differences in protein expression profiles. CyTOF technology is proved feasibility and value at single cell level of gastric cancer.

1. Introduction

Gastric cancer (GC) is the fifth most-common cancer and also the third major cause of cancer associated death across the world [1]. In 2018, it is estimated that 1033,701 new diagnoses and 782,685 deaths from GC reported globally in global today network (http://gco.iarc.fr/ today/data/factsheets/cancers/7-Stomach-fact-sheet.pdf). In addition to surgery, chemotherapy [2] and radiotherapy (RT) [3] are the most important complementary treatments for the treatment of GC.Per-ioperative [4,5] or postoperative chemoradiation [6–8] is the preferred approach for localized GC. To reduce tumor size and inhibit invasion and metastasis before surgery or reduce the risk of recurrence after surgery, neo-adjuvant chemotherapy or RT is given to GC. For un-resectable locally advanced, recurrent or metastatic disease, chemor-adiation is used as a palliative care to reduce suffering and prolong life. Fluorouracil, especially 5-fluorouracil, is one of the main chemother-apeutics for the treatment of GC [9]. Phenotypic and functional het-erogeneity [10] and therapy resistance [11] may be the cause of the difference in the treatment effect or prognosis of chemoradiotherapy, but there is no single cell level study to confirm the relationship

between cell heterogeneit and efficacy. So further research on single cell level is needed.

CyTOF 2 mass cytometer (CyTOF) was developed as a new approach for the detection of proteins at single cell level. CyTOF uniquely com-bined inductively coupled plasma time-of-flight mass spectrometry with maxpar metal-labeling technology. This technique gained un-precedented resolution with 135 detection channels, and more than 40 markers per cell had been measured [12,13].Compared with traditional flow cytometry, CyTOF has the advantage of high multiplicity of bio-marker detection, absolute quantification, absence of detection channel overlap, no sample matrix effects, simplified measurement protocols, and overall lower sample and reagent consumption [13]. CyTOF tech-nology was originally applied to human leukemia, such as immune cell phenotype and function [14], chemotherapy drug efficacy [15,16] and cell phenotype distribution [17]. CyTOF had catalyzed the development of single-cell proteomics, but there have been no reports on the appli-cation of CyTOF technology in GC field.

Here, we analyzed the responses of seventeen proteins to chemor-adiotherapy in human gastric cancer AGS cells with CyTOF technology at single-cell level, and prove applicability of CyTOF technology in GC

https://doi.org/10.1016/j.yexcr.2019.111568

Received 6 May 2019; Received in revised form 17 August 2019; Accepted 20 August 2019

Abbreviations: GC, Gastric cancer; viSNE, Visualization of t-Distributed Stochastic Neighbor Embedding; Citrus, Cluster identification, characterization, and

re-gression; RT, radiotherapy

∗_{Corresponding author. Endoscopy Center of the First Hospital of Hebei Medical University, 050031, Shijiazhuang, Hebei, China.}

E-mail address:ydyynjzx@126.com(W. Yu).

1_{Shuangshuang Han and Xia Jiang contributed equally to this work.}

Experimental Cell Research 384 (2019) 111568

Available online 22 August 2019

(2)

single-cell proteomics. 2. Materials and methods 2.1. Cell line

Human GC AGS cell line was purchased from Culture Collection of the Chinese Academy of Sciences, Shanghai, China. It was derived from fragments of a gastric adenocarcinoma resected from a 54-year-old fe-male Caucasian patient who had received no prior therapy in 1979. It was tumorigenic in athymic BALB/c mice.

2.2. Cell culture

Human GC AGS cells were cultured in RPMI 1640 (Thermo, Grand Island, NY) containing 10% fetal bovine serum (Thermo, Gibco, Grand Island, NY) at 37 °C in a humidified 5% CO2incubator.

2.3. Radiation

Cell cultures were irradiated with single doses, 2·5 Gray (Gy), one time, and the field size at SSD was 30 × 30 cm [18]. An unirradiated control was included as negative control. Forty-eight hours after RT, the expression of seventeen proteins were detected by CyTOF.

2.4. Chemotherapy

Cells were cultured at 37 °C in RPMI 1640 medium supplemented with 10% fetal bovine serum and 5% CO2. Cells were treated with 5-FU (5- fluoroura cil) at a concentration of 1 μg/ml. Then cells were de-tected with CyTOF after 48 h of culturing.

2.5. Mass cytometry

Take 3 million logarithmic growth phase cells after being trypsin-digested and PBS-washed. Centrifuge each sample at 140×g for 5 min (repeated twice), remove the supernatant and collect the cells. Use the above cells for mass cytometry. Add 1 ml methanol into each sample, fix for 10–30 min at −20 °C, mix well with 1 ml of staining buffer, centrifuge at 800×g for 5 min, and remove the supernatant. Add 83 μl staining buffer and antibodies (1 μl/antibody, respectively), mix well and incubate for 30 min at room temperature. Seventeen proteins were randomly selected to assemble a panel with beads, barcodes, DNA test channels and Live/Death test channel, the information of metal con-jugated antibodies were shown in Table 1. Add 1 ml PBS buffer re-spectively and centrifuge each sample at 800×g for 5 min. Add 1 ml PBS respectively, centrifuge each sample at 800×g for 5 min, remove the supernatant. Add 1 mL PBS buffer separately and start the CyTOF2 mass cytometer (Fluidigm Corporation, South San Francisco, CA). Single-cell data were collected in FCS files.

2.6. Data preprocessing

CyTOF data analysis processing was used to find clusters with spe-cific expression combining chemotherapy and RT GC cell lines. CyTOF data was normalized by a MATLAB-based software program called bead normalization [19]. Normalized data was analyzed using cytobank software to generate clean data by eliminating beads and cisplatin-po-sitive events.

2.7. The viSNE implementation

viSNE is a fast, distributed implementation of the t-SNE algorithm, improved and tailored for the analysis of single-cell data [20]. viSNE is a computational tool that projects cells onto a two dimensional map such that the distances between cells in 2D reflect the distance between

them in high-dimensional space [21]. Thus, cells that are close together on a viSNE map are phenotypically similar for the markers used to create the map. Users then identify and characterize populations of cells based on the groups formed in the viSNE map. We performed a viSNE analysis of the GC cell lines among 5-FU Group, RT Group and Con Group. In viSNE, a point in high dimensional space represents each cell is, and each coordinate representing one measured parameter (e.g., the protein expression level), the expression intensity of each parameter can be visualized as heat intensity on the viSNE map, enabling identi-fication of differently expressed cells between different groups. Im-portantly, viSNE allowed a global single-cell view of the data. 2.8. CITRUS analysis

Given cytometry data from many samples and an endpoint of interest for each sample (e.g., good or poor patient outcome, patient survival time), Citrus identifies clusters of phenotypically similar cells in an un-supervised manner, characterizes the behavior of identified clusters by using biologically interpretable metrics, and leverages regularized su-pervised learning algorithms to identify the subset of clusters whose behavior is predictive of a sample's endpoint. Citrus (cluster identifica-tion, characterization and regression) was developed as a data-driven approach for the identification of stratifying subpopulations in multi-dimensional cytometry datasets. The False Discovery Rate (FDR) threshold helps to make sure that models produced by PAM don't include too many false positive features. For any given Citrus run, a trait will either be the median intensity of a biological marker in that cluster, depending on the setup parameters of the run. Each data point on the model error rate graph is a model that Citrus is evaluating. The number of features in each model is shown at the top of the graph. The cross-validation error rate is the percentage of times that the model is wrong at predicting the sample group that a file belongs to when it is not included in the training set to build a model ( https://support.cytobank.org/hc/en-us/articles/226678087-How-to-Configure-and-Run-a-CITRUS-Analysis).

Table 1

The information of metal conjugated antibodies panel.

Channels Name Metal Parameters/antibodies

Pd102Di Pd102Di Barcode

Ce140Di Ce140Di Barcode

Pr141Di Pr141Di CD326

Nd142Di Nd142Di Beads

Ce142Di Ce142Di Beads

Nd144Di Nd144Di IL-4

Nd148Di Nd148Di GAL

Nd150Di Nd150Di MIP-1

Eu151Di Eu151Di IL-5

Sm152Di Sm152Di TNFα

Eu153Di Eu153Di Beads

Sm154Di Sm154Di CD45

Gd156Di Gd156Di IL-6

Gd158Di Gd158Di IL-2

Gd160Di Gd160Di CD14

Dy164Di Dy164Di IL-17

Ho165Di Ho165Di Beads

Er168Di Er168Di IFNg

Er170Di Er170Di CD3

Yb171Di Yb171Di GranzymB

Yb173Di Yb173Di CD104

Yb174Di Yb174Di CK8/18

Lu175Di Lu175Di Perforin

Ir191Di Ir191Di DNA

Ir193Di Ir193Di DNA

Pt195Di Pt195Di Live/Dead

(3)

The minimum cluster-size threshold (MCST) is set up 10%.Table 1shows a mass cytometry staining panel.

2.9. Statistical analysis

All results were shown as mean ± SD and analyzed using GraphPad Prism 7 (GraphPad Software, San Diego, California) from at least three independent experiments. The differences between the groups were calculated with Mann-Whitney test. Data were considered statistically significant when P < 0·05.

3. Results

3.1. Apoptosis of AGS cells were induced by chemoradiotherapy

Cells were treated by 1 μg/ml 5-FU or 2·5 Gy RT, the cells were col-lected and stained by metal conjugated antibodies (antibodies informa-tion shown inTable 1) at 48 h after the treatment. The size of AGS cells was measured, approximately 18–45 μm, and the cells could pass through nebulizer and quadrupole of CyTOF machine, metal isotopes were detected by CyTOF, analysis of flow chart was shown inFig. 1.And cellular apoptosis was analyzed by Cisplatin (Pt 195 Pi) through Cyto-bank. Compared with Con Group cell counts (45491 ± 1385·1), apop-tosis was induced in 5-FU and RT Group (58724 ± 620·6, 49363 ± 1097·7, P < 0.05, respectively). Chemoradiotherapy induced cell apoptosis was also verified by traditional method, cell counting kit-8 (CCK-8) assay (Fig. S5) and colony formation assay (Fig. S6)

3.2. Phenotypes of AGS cells revealed by ViSNE after chemoradiotherapy The expression profiles of AGS cells were visualized by viSNE. viSNE plots revealed that 5-FU chemotherapy and RT induced some changes in the phenotype and function of AGS cells. AsFig. 2shown, the col-orful plots showed that 5-FU-treated and radiation-induced cells loca-lized in the different areas of the viSNE plot, demonstrating that the combinatorial marker expression changes of AGS cells after 5-FU che-motherapy or RT. Specifically, 5-FU cheche-motherapy decreased cell density in some makers’ blue areas of the plot, for example, IL-6, TNFα, IFNg, CK8/18 and GranzymB. Phenotypic changes in other 12 proteins after chemoradiotherapy were seen inFig. S1. To prove the feasibility of the viSNE method, MGC-803 (another GC cell line) cells were de-tected and analyzed by the same method, and data was shown in Fig. S7.

3.3. Automatically subset single cell data was calculated by citrus among groups of AGS cells

Furthermore, we used Citrus to undertake a more systematic in-vestigation of phenotypic and functional differences between samples after 5-FU chemotherapy and RT [22]. Citrus constructed classification models using significance analysis of microarrays (SAM) and nearest shrunken centroid (PAMR) methods. We chose nearest shrunken cen-troid method. Plotting the classification error rates and feature false discovery rates (FDRs) of each model were estimated with Cross vali-dation and permutation tests respectively (Fig. 3A). The corresponding

(4)

phenotypes for clusters whose behavior differs between the three sti-mulation groups were described. Citrus divided AGS cells into 16 clusters and automatically picked out the noteworthy clusters by models running in 10% minimum cluster size, 1 cross-validation, and 1% false discover rate.

Automatically subsets were shown inFig. 3B. The clusters of worthy attention were colored to wine red in light gray area, including 44992, 44993, 44994, 44996 and 44999 clusters in single cell data of CD326 expression. The worthy clusters of other proteins expression single cell data were shown inFig. S2.

Then, all clusters data were analyzed and found two specific clus-ters, 44992 and 44993 clusclus-ters, which had different protein expression profiles within each group. Cluster 44992 had a phenotype of low ex-pression of TNFα, IL-6, IFNg, GranzymB and CK8/18 while cluster 44993 expression profile was nearly converse to cluster 44992, which role out a high expression of TNFα, IL-6, IFNg, GranzymB and no ex-pression change of CK8/18 (Fig. 3C). The expression profiles of other worthy clusters were shown inFig. S3.

Further analyzed different expression levels of seventeen proteins in 44992, 44993 and other worthy clusters cells among control, 5-FU, and RT group. As shown in Fig. 3D, compared with Con Group, CD326 expression was upregulated by 5-FU and RT treatment in the both clusters. The expression of GAL, IL-4 and IL-17 was increased by 5-FU in 44992 cluster cells, however 5-FU induced the expression of IL-5 and CD45 in 44993 cluster cells. The expression levels of other proteins were shown inFig. S4.

To prove the feasibility of the Citrus method, the FDR (Fig. S8A), Citrus network (Fig. S8B), Citrus histograms (Fig. S8C) and different protein expression profiles (Fig. S8D) were analyzed by Citrus in MGC-803 cells. Two specific clusters, 1473 and 1484, were found that MIP-1 expression was upregulated by 5-FU and RT treatment in 1473 cluster, and GAL expression was decreased by 5-FU and RT treatment in 1484 cluster. This result proved that Citrus analysis method was feasibility in single cell level study.

3.4. The overall expression level was changed in 5-FU chemotherapy and RT treated AGS cells

In AGS cells, Mann-Whitney test showed that the expression of TNFα, IL-6, IFNg, GranzymB and CK8/18 in 5-FU Group was higher than that in Con Group, the differences were statistically significant (P < 0·05). Combined with the median expression of each protein in Cytobank, the different expression proteins (> 1·5 fold change) in-cluded TNFα, IL-6, and GranzymB.

While the expression of TNFα, IL-6, IFNg, GranzymB and CK8/18 in RT Group was higher than that in Con Group, the differences were statistically significant (P < 0·05). Combined with the median ex-pression of each protein in Cytobank, there was no different exex-pression protein (> 1·5 fold change,Table 2).

4. Discussion

The data of single-cell analysis has opened a new area for analyzing cancer cell heterogeneity. In genomics, single-cell sequencing technique has ushered a development of genetic evolution of cancer, including genetic mutation and copy number variation [23]. In proteomics, the techniques were almost based on flow cytometry and mass spectro-scopy, including flow cytometry, CyTOF, single-cell proteomics by mass-spectrometry (SCoPE-MS) [24]. Modern flow cytometry can per-form simultaneous detection of 10–20 parameters per cell, at throughput rates above 10,000 cells per second [25]. State-of-the-art systems may reach as many as 50 parameters [26]. Mass cytometry has the characteristics of high-speed analysis of traditional flow cytometry and high-resolution capability for mass spectrometry. ICP-MS device has a very wide atomic weight detection range (88–210) which can be carried out multiple detections, it can measure hundreds of different parameters simultaneously with no interference in each channel. In addition, due to the non-specific binding ability of metal tags and cel-lular components, its background signal is extremely low [27]. Ex-tensive and dynamical protein modulation of tumors may be more critical than RNA, which becomes a rising concerning [28]. At the present study, the heterogeneity was first detected by CyTOF in GC AGS cells with or without chemoradiotherapy.

In this study, the size of AGS cells was measured, approximately 18–45 μm, and the cells could pass through nebulizer and quadrupole of CyTOF machine. Then, the panel was selected (Table 1), including se-venteen proteins, depending on the main panels of SciLifeLab National Mass Cytometry Facility at Linkoping University (http://cytof. scilifelab.se/panels). We analyzed the responses of seventeen proteins to chemoradiotherapy by viSNE and Citrus at single-cell level in human GC AGS cells. Two specific clusters are found, which had differences in the protein expression profiles (Figs. 3B and S3). This study verified applicability of CyTOF technology at the single cell level of GC.

The heterogeneity of protein expression profiles was shown in 44992 and 44993 clusters. The expression of TNFα, IL-6, IFNg, GranzymB and CK8/18 were different in 44992 and 44993 clusters. Cluster 44992 cells had the low expression of TNFα, IL-6, IFNg, GranzymB and CK8/18, while cluster 44993 cells showed a nearly

Fig. 2. viSNE-displayed the expression

changes of combinatorial markers (IL-6, TNFα, IFNg, CK8/18 and GranzymB) on AGS cells after 5-FU chemotherapy or radiotherapy. T-stochastic neighborhood embedding (t-SNE) plots shows that 5-FU-treated and radiation-induced cells loca-lized in the different areas of the viSNE plot. Expression plots are represented in a color scale, from blue (low) to red (high).

(5)

converse phenotype to cluster 44992 which had the high expression of TNFα, IL-6, IFNg, Granzym B and no expression change of CK8/18. It has been reported that the aberrant productions of TNFα and IL-6 were associated with treatment resistance [29–32]. Thus, we speculate that 44992 cluster cells may be resistant to chemotherapy or RT, compared with 44993 cluster cells. More study is needed to investigate the effect of the different protein expression profiles of 44992 or 44993 clusters in the future, through cell separation from AGS cells. We also tested an-other GC cell line MGC803 in the same way and confirmed the feasi-bility of the viSNE and Citrus methods, results proved that methods was feasibility in single cell level study (Figs. S7–S8).

In the future, we will re-assemble panels, which includes epidermal growth factor receptor (EGFR, for anti-EGFR durgs cetuximab and pa-nitumumab screening), vascular endothelial growth factor/vascular

endothelial growth factor receptor (VEGF/VEGFR, for anti-VEGFR durgs bevacizumab and apatinib screening), human epidermal growth factor receptor 2 (HER2, for anti-HER2 durgs trastuzumab and lapatinib screening), and apply CyTOF technology to evaluate the therapeutic sensitivity combined with traditional chemotherapy drugs (docetaxel, oxaliplatin, cisplatin etc.), and further guide individualized medication. Our study demonstrates the applicability of CyTOF in GC cells and the heterogeneity of protein expression profiles was verified. Two specific clusters are found which had differences in protein expression profiles, which may be the cause of the difference in the treatment ef-fect or prognosis of chemoradiotherapy. Depending on detection ad-vantage (up to 135 parameters), CyTOF will bring about breakthrough at single-cell level in GC research field, through designing new panels.

Fig. 3. Automatically subset single cell data was calculated by Citrus among groups of AGS cells. (A) The False Discovery Rate (FDR) threshold helps to make sure

that models produced by PAM don't include too many false positive features. Each data point on the model error rate graph is a model that Citrus is evaluating. Our analysis produces models with low (good) cross validation error rates. (B) Citrus network tree visualizing clusters of worthy attention colored to wine red in light gray area, including 44992, 44993, 44994, 44996 and 44999 clusters in single cell data of CD326 expression. Circle size reflects number of cells within a given cluster. (C) Two specific clusters, 44992 and 44993 clusters, which had different protein expression profiles within each group generated by Citrus histograms. Cluster 44992 has a phenotype of low expression of TNFα, IL-6, IFNg, GranzymB and CK8/18 while cluster 44993 shares a nearly converse phenotype to cluster 44992 which role out a high expression of TNFα, IL-6, IFNg, GranzymB and no expression change of CK8/18. (D) Further analyzed different expression levels of seventeen proteins in 44992, 44993 among Con, 5-FU, and RT groups. Compared with Con Group, CD326 expression was upregulated by 5-FU and RT treatment in both clusters. The expression of GAL, IL-4 and IL-17 was increased by 5-FU in 44992 cluster cells, however 5-FU induced IL-5 and CD45 expression in 44993 cluster cells.

(6)

Funding sources

This work was supported by Hebei Province Key Research and Development Project (18277741D), International Science & Technology Cooperation Program of China (2014DFA31150) and other Hebei Province Projects (1387, SGH201501, A201802017, LNB201809 and G2019035).

Declarations of interests

The authors declare no conflicts of interest. Author contributions

Conception and design of study: Weifang Yu, Xiao-Feng Sun, Hong Zhang, Zengren Zhao

Performed the experiments: Xia Jiang, Shuangshuang Han, Jan-Ingvar Jönsson.

Data analysis and interpretation: Shuangshuang Han, Xia Jiang. Manuscript writing: Shuangshuang Han, Xia Jiang, Chao Li. Project supervision: Xiao-Feng Sun, Weifang Yu, Zengren-Zhao. Final approval of manuscript: All authors.

Acknowledgments

We thank Prof. Jan-Ingvar Jönsson, Dr. Mikael Pihl and Dr. Florence Sjogren of Linkoping University in Sweden to provide subject guidance and CyTOF equipment.

Appendix A. Supplementary data

Supplementary data to this article can be found online athttps:// doi.org/10.1016/j.yexcr.2019.111568.

References

[1] K. Wang, Q. Liang, X. Li, H. Tsoi, J. Zhang, H. Wang, M.Y. Go, P.W. Chiu, E.K. Ng, J.J. Sung, J. Yu, MDGA2 is a novel tumour suppressor cooperating with DMAP1 in gastric cancer and is associated with disease outcome, Gut 65 (2016) 1619–1631. [2] N. Kothari, K. Almhanna, Current status of novel agents in advanced

gastro-esophageal adenocarcinoma, J. Gastrointest. Oncol. 6 (2015) 60–74.

[3] N. Zhang, Q. Fei, J. Gu, L. Yin, X. He, Progress of preoperative and postoperative radiotherapy in gastric cancer, World J. Surg. Oncol. 16 (2018) 187.

[4] M. Ychou, V. Boige, J.P. Pignon, T. Conroy, O. Bouche, G. Lebreton, M. Ducourtieux, L. Bedenne, J.M. Fabre, B. Saint-Aubert, J. Geneve, P. Lasser, P. Rougier, Perioperative chemotherapy compared with surgery alone for resectable

gastroesophageal adenocarcinoma: an FNCLCC and FFCD multicenter phase III trial, J. Clin. Oncol. : Off. J. Am. Soc. Clin. Oncol. 29 (2011) 1715–1721. [5] D. Cunningham, W.H. Allum, S.P. Stenning, J.N. Thompson, C.J. Van de Velde,

M. Nicolson, J.H. Scarffe, F.J. Lofts, S.J. Falk, T.J. Iveson, D.B. Smith, R.E. Langley, M. Verma, S. Weeden, Y.J. Chua, M.T. Participants, Perioperative chemotherapy versus surgery alone for resectable gastroesophageal cancer, N. Engl. J. Med. 355 (2006) 11–20.

[6] S.R. Smalley, J.K. Benedetti, D.G. Haller, S.A. Hundahl, N.C. Estes, J.A. Ajani, L.L. Gunderson, B. Goldman, J.A. Martenson, J.M. Jessup, G.N. Stemmermann, C.D. Blanke, J.S. Macdonald, Updated analysis of SWOG-directed intergroup study 0116: a phase III trial of adjuvant radiochemotherapy versus observation after curative gastric cancer resection, J. Clin. Oncol. : Off. J. Am. Soc. Clin. Oncol. 30 (2012) 2327–2333.

[7] X. Wang, D.B. Zhao, L. Yang, Y. Chi, Y. Tang, N. Li, S.L. Wang, Y.W. Song, Y.P. Liu, W.Y. Liu, H. Ren, T. Zhang, J.Y. Wang, X.S. Chen, H. Fang, W.H. Wang, Y.X. Li, J. Jin, S-1 chemotherapy and intensity-modulated radiotherapy after D1/D2 lymph node dissection in patients with node-positive gastric cancer: a phase I/II study, Br. J. Canc. 118 (2018) 338–343.

[8] J. Ng, P. Lee, The role of radiotherapy in localized esophageal and gastric cancer, Hematol. Oncol. Clin. N. Am. 31 (2017) 453–468.

[9] H. Qiu, Z. Zhou, [Updates and interpretation on NCCN clinical practice guidelines for gastric cancer 2017 version 5], Zhonghua wei chang wai ke za zhi = Chin. J. Gastrointest. Surg. 21 (2018) 160–164.

[10] C.E. Meacham, S.J. Morrison, Tumour heterogeneity and cancer cell plasticity, Nature 501 (2013) 328–337.

[11] S.V. Sharma, D.Y. Lee, B. Li, M.P. Quinlan, F. Takahashi, S. Maheswaran, U. McDermott, N. Azizian, L. Zou, M.A. Fischbach, K.K. Wong, K. Brandstetter, B. Wittner, S. Ramaswamy, M. Classon, J. Settleman, A chromatin-mediated re-versible drug-tolerant state in cancer cell subpopulations, Cell 141 (2010) 69–80. [12] D.R. Bandura, V.I. Baranov, O.I. Ornatsky, A. Antonov, R. Kinach, X. Lou, S. Pavlov,

S. Vorobiev, J.E. Dick, S.D. Tanner, Mass cytometry: technique for real time single cell multitarget immunoassay based on inductively coupled plasma time-of-flight mass spectrometry, Anal. Chem. 81 (2009) 6813–6822.

[13] O. Ornatsky, D. Bandura, V. Baranov, M. Nitz, M.A. Winnik, S. Tanner, Highly multiparametric analysis by mass cytometry, J. Immunol. Methods 361 (2010) 1–20.

[14] A.J. Lamble, M. Dietz, T. Laderas, S. McWeeney, E.F. Lind, Integrated functional and mass spectrometry-based flow cytometric phenotyping to describe the immune microenvironment in acute myeloid leukemia, J. Immunol. Methods 453 (2018) 44–52.

[15] M. Deng, J. Zha, Z. Jiang, X. Jia, Y. Shi, P. Li, X.L. Chen, Z. Fang, Z. Du, B. Xu, Apatinib exhibits anti-leukemia activity in preclinical models of acute lympho-blastic leukemia, 16 (2018) 47.

[16] B.Z. Carter, P.Y. Mak, H. Mu, H. Zhou, D.H. Mak, W. Schober, J.D. Leverson, B. Zhang, R. Bhatia, X. Huang, J. Cortes, H. Kantarjian, M. Konopleva, M. Andreeff, Combined targeting of BCL-2 and BCR-ABL tyrosine kinase eradicates chronic myeloid leukemia stem cells, Sci. Transl. Med. 8 (2016) 355ra117. [17] L. Han, P. Qiu, Z. Zeng, J.L. Jorgensen, D.H. Mak, J.K. Burks, W. Schober,

T.J. McQueen, J. Cortes, S.D. Tanner, G.J. Roboz, H.M. Kantarjian, S.M. Kornblau, M.L. Guzman, M. Andreeff, M. Konopleva, Single-cell mass cytometry reveals in-tracellular survival/proliferative signaling in FLT3-ITD-mutated AML stem/pro-genitor cells, Cytometry, Part A : J. Int. Soc. Anal. Cytol. 87 (2015) 346–356. [18] S. Gnosa, H. Zhang, V.P. Brodin, J. Carstensen, G. Adell, X.F. Sun, AEG-1 expression

is an independent prognostic factor in rectal cancer patients with preoperative radiotherapy: a study in a Swedish clinical trial, Br. J. Canc. 111 (2014) 166–173. [19] R. Finck, E.F. Simonds, A. Jager, S. Krishnaswamy, K. Sachs, W. Fantl, D. Pe'er,

G.P. Nolan, S.C. Bendall, Normalization of mass cytometry data with bead stan-dards, Cytometry Part A : the journal of the International Society for Analytical Cytology 83 (2013) 483–494.

[20] A.D. Amir el, K.L. Davis, M.D. Tadmor, E.F. Simonds, J.H. Levine, S.C. Bendall, D.K. Shenfeld, S. Krishnaswamy, G.P. Nolan, D. Pe'er, viSNE enables visualization of high dimensional single-cell data and reveals phenotypic heterogeneity of leu-kemia, Nat. Biotechnol. 31 (2013) 545–552.

[21] B. Becher, A. Schlitzer, J. Chen, F. Mair, H.R. Sumatoh, K.W. Teng, D. Low, C. Ruedl, P. Riccardi-Castagnoli, M. Poidinger, High-Dimens. Anal. Murine Myeloid Cell Syst. 15 (2014) 1181–1189.

[22] R.V. Bruggner, B. Bodenmiller, D.L. Dill, R.J. Tibshirani, G.P. Nolan, Automated identification of stratifying signatures in cellular subpopulations, Proc. Natl. Acad. Sci. U. S. A 111 (2014) E2770–E2777.

[23] X. Zhang, S.L. Marjani, Z. Hu, S.M. Weissman, X. Pan, S. Wu, Single-cell sequencing for precise cancer research: progress and prospects, Cancer Res. 76 (2016) 1305–1312.

[24] B. Budnik, E. Levy, G. Harmange, N. Slavov, SCoPE-MS: Mass Spectrometry of Single Mammalian Cells Quantifies Proteome Heterogeneity during Cell Differentiation vol. 19, (2018), p. 161.

[25] L.M. Weber, M.D. Robinson, Comparison of clustering methods for high-dimen-sional single-cell flow and mass cytometry data, Cytometry Part A : the journal of the International Society for Analytical Cytology 89 (2016) 1084–1096. [26] Y. Saeys, S. Van Gassen, B.N. Lambrecht, Computational flow cytometry: helping to

make sense of high-dimensional immunology data, Nat. Rev. Immunol. 16 (2016) 449–462.

[27] S.C. Bendall, G.P. Nolan, M. Roederer, P.K. Chattopadhyay, A deep profiler's guide to cytometry, Trends Immunol. 33 (2012) 323–332.

[28] C. Vaklavas, S.W. Blume, W.E. Grizzle, Translational dysregulation in cancer: mo-lecular insights and potential clinical applications in biomarker development, Front. Oncol. 7 (2017) 158.

Table 2

Median expression of each indicator in Cytobank in AGS cells.

Antigens 5-FU RT Con

IL-5 4.99 2.41 2.51 CD45 6.03 4.34 4.06 IL-6 4.96 2.77 2.77 Perforin 16.23 10.42 10.66 Galectin 109 81.6 85.5 TNFα 6.76 3.55 3.95 CD326 2434 2140 1691 IL-4 17 11.3 11.7 IL-2 7.09 4.06 4.24 CD14 4.24 3.04 3 IL-17 13.07 6.6 7.9 IFNg 13.4 7.74 9.43 CD3 2.3 1.4 1.3 CK8/18 527.14 347.2 355.46 MIP-1 5.6 3.74 3.61 GranzymB 4.26 2.36 2.66 CD104 36.47 43.76 29.39

Note: 5-FU Group (AGS cells dealt with 5- fluorouracil), RT Group (AGS cells dealt with RT) and Con Group (AGS cells that have not been treated).

(7)

[29] T. Wu, Y. Dai, Tumor microenvironment and therapeutic response, Cancer Lett. 387 (2017) 61–68.

[30] T. Rabi, S. Shukla, S. Gupta, Betulinic acid suppresses constitutive and TNFalpha-induced NF-kappaB activation and induces apoptosis in human prostate carcinoma PC-3 cells, Mol. Carcinog. 47 (2008) 964–973.

[31] E. Stanzani, F. Martinez-Soler, T.M. Mateos, N. Vidal, A. Villanueva, M.A. Pujana, J. Serra-Musach, N. de la Iglesia, P. Gimenez-Bonafe, A. Tortosa, Radioresistance of mesenchymal glioblastoma initiating cells correlates with patient outcome and is

associated with activation of inflammatory program, Oncotarget 8 (2017) 73640–73653.

[32] P. Sansone, C. Ceccarelli, M. Berishaj, Q. Chang, V.K. Rajasekhar, F. Perna, R.L. Bowman, M. Vidone, L. Daly, J. Nnoli, D. Santini, M. Taffurelli, N.N. Shih, M. Feldman, J.J. Mao, C. Colameco, J. Chen, A. DeMichele, N. Fabbri, J.H. Healey, M. Cricca, G. Gasparre, D. Lyden, M. Bonafe, J. Bromberg, Self-renewal of CD133(hi) cells by IL6/Notch3 signalling regulates endocrine resistance in meta-static breast cancer, Nat. Commun. 7 (2016) 10442.